The invention relates to a method to operate a gesture recognition system for a motor vehicle as well as a gesture recognition system.
In modern motor vehicles, gesture recognition systems are increasingly used for the control of additional functions of the motor vehicle. A plurality of different gesture recognition systems are applied here. These are both contactless gesture recognition systems which, for example, recognize a gesture via a camera, and gesture recognition systems which must be touched by a user in order to be able to recognize a gesture. Examples here are touch-sensitive screens or also optical finger navigation modules which are often also referred to as “OFN”. It applies for all gesture recognition systems in a motor vehicle that a speed, above all however a direction, of a gesture is to be defined as precisely as possible. The current speed and the direction of a gesture can be depicted, i.e., represented, by a velocity vector. The direction thereof then represents the direction of the detected gesture and its length represents the speed of the gesture. A velocity vector is then often classified in the scope of a quantification of the gesture to determine the direction of a gesture, in order, for example, to differentiate horizontal and vertical gestures from each other.
In U.S. Pat. No. 8,212,794 B2, an optical finger navigation module is disclosed which is designed to quantify and to classify operating gestures by means of a comparison of detected movement information with linear limit functions.
The object of the present invention is to provide a method to operate a gesture recognition system for a motor vehicle, with which recognition reliability as well as classification reliability is increased. The object of the invention is also to provide a device to implement such a method.
This object is solved according to the invention by a method or a gesture recognition system according to the independent claims. Advantageous embodiments of the invention are subject matter of the dependent claims, the description and the figures.
in order to increase recognition reliability and classification reliability, a series of steps are provided in a method according to the invention to operate a gesture recognition system for a motor vehicle: Firstly, a detection of a movement gesture which belongs to an operating action, by a detection unit. Then a generation of a movement vector, which represents the detected gesture, by the detection unit, followed by a transmission of the movement vector to a processing unit. Then a comparison of the movement vector with at least two limit curves, which define at least three separate regions in a section of a vector space, by the processing unit. Then, depending on a result of the previously implemented comparison, a determination of a direction of the movement gesture by the processing unit as horizontal if the movement vector lies in a first region of the three separate regions, as vertical if the movement vector lies in a second region of the three separate regions, and as not able to be classified if the movement vector lies in a third region of the three separate regions. It is essential for the invention here that the limit curves be defined by functions which depend on the spatial directions in a non-linear manner.
The movement gesture is therefore quantized here in a non-linear way into a mainly horizontal or vertical movement. This has the advantage that a limited sensor resolution can be considered via a design of non-linearity, and therefore, especially for slow finger movements, in the case of which in general the directional information is less accurate due to the limited sensor resolution than for quick finger movements, increased recognition reliability is achieved. Due to the non-linear limit curves, the directional tolerance for slow movements can in particular be designed to be greater than for quick movements. Furthermore, typical movement characteristics which occur during the operation of the gesture recognition system can also be considered during gesture classification or quantization via the non-linear functions, and therefore the classification reliability can be increased.
In a preferred embodiment of the invention, it is provided that the spatial directions, on which the functions which define the limit curves depend, are an x-direction and a y-direction. The limit curves are therefore defined in a plane. This has the advantage that the method increases recognition reliability and classification reliability for two-dimensional movement gestures. In particular, scrolling as an operating action can therefore be processed efficiently. Additionally, a two-dimensional implementation of the method is simple to achieve as only a low calculation capacity is required.
In a particularly advantageous embodiment, it is provided that the two non-linear functions, which define the limit curves if they are transformed into a vector angle-vector length space, assume a finite, predetermined vector angle initial value, in particular between 30° and 60°, for infinitely small vector lengths and a finite, predetermined vector angle final value, in particular between 10° and 80°, for infinitely large vector lengths. This has the advantage that a classification range of the movement gesture can therefore be defined explicitly for short vectors, therefore slow movements, and at the same time a limit range can be defined in this way for long vectors, i.e., quick movements, by no classification occurring. As the two functions are non-linear, the two advantages referred can be adjusted explicitly and independently of each other, and furthermore almost any characteristics of the classification or quantization of the gesture movement can furthermore be transferred between the extremes of very slow and very quick movements, so very short and very long vectors.
In a preferred embodiment, it is provided that the vector angle final value of the first non-linear function is greater than its vector angle initial value and the vector angle final value of the second non-linear function is smaller than its vector angle initial value, wherein the vector angle initial value of the first non-linear function is greater than or equal to the vector angle initial value of the second non-linear function. This leads to the limit curves for small vector lengths being closer to each other in the vector angle-vector length space than for large vector lengths. Therefore in particular the angle range of the vector space in which a vector is not allocated to a horizontal or vertical movement is classified as smaller for small vector lengths than for large vector lengths. This has the advantage that the directional tolerance for slow movements is greater than for quick movements, so in particular deviations from the horizontal and/or vertical direction for slow movement gestures lead to a determination of the direction of the movement gesture as “not able to be classified” less easily than for quick movement gestures. Therefore, recognition reliability and classification reliability are increased, especially for slow movements.
It can furthermore be provided that, before the comparison of the vectors with the limit curves, negative components of the movement vector are multiplied by −1 and the section of the vector space in which the comparison is carried out exclusively comprises values greater than or equal to zero. In particular, the section of the vector space therefore comprises the so-called first quadrant which is defined as greater than or equal to zero in an x-y-plane by x- and y-values. In other words, for a classification of the movement gesture, for example, movement gestures directed to the left or to the right therefore cannot be differentiated between, but only the angle of the movement gesture is regarded. This has the advantage that the limit curves can be provided particularly simply and an expensive checking of applicable ranges of different limit functions is not required.
Analogously, before the comparison, the limit curves can also be mirrored from the first quadrant into the other quadrants and the components of the movement vector can be left unchanged. The comparison then therefore occurs in the entire vector space, but outside the first quadrant, only with the respective mirror images of the limit curves and not with the functions from which the limit curves are derived.
The invention also relates to a gesture recognition system for a motor vehicle. This has a detection unit which is designed to detect a movement gesture which belongs to an operating action, to generate a movement vector which represents the movement gesture and to transmit the movement vector to a processing unit. This processing unit is likewise part of the gesture recognition system and is designed to compare the movement vector with at least two limit curves which define at least three separate regions in a section of a vector space and, depending on a result of the comparison, to determine a direction of the movement gesture as horizontal if the movement vector lies in a first region of the three separate regions, as vertical if the movement vector lies in a second region of the three separate regions and as not able to be classified if the movement vector lies in a third region of the three separate regions. Herein the processing unit is designed to use limit curves for the comparison which are defined by functions which depend on spatial directions in a non-linear manner.
It can be provided that the detection unit is a touch-sensitive detection unit. This has the advantage that an efficient determination of the velocity vector is possible as, especially in a motor vehicle having several passengers, the movement data is fundamentally only detected if in fact a touching is also present, so a gesture is intended and therefore a gesture recognition is desired.
In a particularly advantageous embodiment it is also provided that the detection unit is an optical finger navigation module. Such modules, also referred to as “OFN”, are suitable in a very specific way for an application of the method, as these typically provide less accurate directional information especially for slow finger movements, than for quick finger movements. Therefore, increased recognition reliability or classification reliability is particularly desirable here. In particular, such modules are particularly small, applicable in a varied manner and save electricity, such that they are expedient for use in a motor vehicle.
The preferred embodiment presented with regard to the method according to the invention and the advantages thereof apply accordingly for the device according to the invention and vice versa,
Further features of the invention result from the claims, the figures and the description of the figures. All features and feature combinations referred to above in the description as well as the features and feature combinations referred to below in the description of the figures and/or shown only in the figures are not only applicable in the respectively specified combination, but also in other combinations or, however, alone.
Exemplary embodiments of the invention are explained below by means of the figures.